Claude Content Skills

Claude content skills are editorial procedures written down once: how an article is structured in your project, what separates newsletter copy from landing page copy, which tone is acceptable and which is not. Claude applies them identically on every run, so the output stops depending on how well someone happened to phrase the request that day.

The hard problem in generated text is not model quality but reproducibility. The same request asked twice returns different structure and tone, and an editor spends the saved time reconciling them. A skill pins the structure and the style rules to a file and removes that variance.

Editorial policy is a separate win. Prohibitions — never promise results, never name competitors, no figure without a source — live in the skill rather than in an editor's memory, and they apply to every piece, including the ones written in a hurry.

The collection below covers drafting and editing, marketing copy, and publication prep.

Skills in this collection

• Сторителлинг на основе данных — data-storytelling is a Claude Code skill that transforms raw data into compelling narratives designed to drive decisions and inspire action. It provides a six-step narrative arc — Hook, Context, Rising Action, Climax, Resolution, and Call to Action — alongside a Setup/Conflict/Resolution framework and a three-pillar model covering data, narrative, and visuals. Extended pattern documentation and worked examples are stored in `references/details.md`. The skill is built for analysts, product managers, and business leaders who need to craft executive presentations, quarterly business reviews, investor decks, or data-driven reports for non-technical audiences — any situation where the goal is not just to surface numbers but to communicate insight and drive a clear next step.
• Анализ вирусности контента через теории коммуникации — dbs-spread is a Claude Code skill that decodes the psychological resonance mechanisms behind content spread using five classical communication theories. It runs the submitted text through Spiral of Silence (Noelle-Neumann, 1974), Uses and Gratifications (Katz, 1959), Framing Theory (Entman, 1993), Two-Step Flow (Lazarsfeld, 1944), and Cognitive Dissonance Theory (Festinger, 1957), producing a one-line finding plus a theoretical rationale for each dimension. The skill then synthesizes a structured output covering the audience's core emotional layer, the effective stance the content takes, what topics the audience still wants to hear on the same emotional frequency, and a focused discussion direction for community chat. It does not predict platform performance metrics or rate content quality. Marketers, content strategists, and community managers use it to understand why certain content travels and how to deepen engagement in the same emotional register.
• Редактура сценария для удержания зрителя — dbs-script-flow is a Claude Code skill that checks short-video scripts for paragraph cohesion, information density, and spoken-word fluency, pinpointing the exact moments where viewers are likely to swipe away. Analysis runs across three dimensions: logical continuity between paragraphs, redundancy or filler within each segment, and sentence-level readability for delivery out loud. Every issue is tagged with a risk level — 🔴 high (likely drop-off), 🟡 medium (possible distraction), 🟢 minor suggestion — so creators can prioritize fixes. Once the audit is complete, the skill offers in-line markup editing: the original text is preserved with strikethroughs alongside 🆕 replacements, making every change immediately visible. It is built for short-form video creators who want to improve completion rates and polish voice-over scripts while keeping their own tone intact.
• Цветовая система для цифровых продуктов — better-colors is a Claude Code skill that covers the full lifecycle of a color system for digital products — from building and naming a palette to applying it consistently and verifying contrast. Six reference files handle distinct concerns: palette structure (ramps, neutrals, status colors), ramp generation from a brand color with dark-mode support, two-tier token naming (primitives by hue, semantic tokens by role), color usage rules (one meaning per color, gradients, emphasis), APCA and WCAG contrast checking, and format work including notation choice, conversion, and Display P3 fallbacks. Contrast accessibility is intentionally delegated to better-accessibility; surfaces and shadows belong to better-ui. The skill suits teams creating a new oklch-based system, auditing an existing codebase, or migrating to semantic design tokens.
• Оптимизация хука для коротких видео — dbs-hook is a Claude Code skill that diagnoses problems in the first five seconds of short-form video and generates optimized hook options for TikTok, Instagram Reels, and YouTube Shorts. It follows a five-phase workflow: first assessing content quality and material richness across five dimensions — impactful data, transformation stories, memorable quotes, authority references, and pain-point resonance — before proceeding to any hook optimization. If the content passes, the skill audits the existing hook against six criteria: independence from the title, presence of an attention grab, suspense without revealing the answer, credibility, voiceover friendliness, and alignment with the body content. It then produces 10–15 alternative openings using three methods: extracting existing material, supplementing with additional facts, and reframing conclusions as open questions. Designed for creators who want to reduce early drop-off rates and improve audience retention on short-video platforms.
• Принципы полировки UI-компонентов — better-ui is a Claude Code skill that provides design engineering principles for making interfaces feel polished and considered. It covers 13 concrete principles with exact values: concentric border radius (outer = inner + padding), optical over geometric alignment, layered box-shadows for depth, interruptible CSS transitions, stagger entrances at ~100 ms intervals, scale(0.96) on press, icon cross-fading via opacity/scale/blur (0.25→1, 4px→0px), and transition suppression during theme switches. The skill organises guidance across surfaces, enter/exit animations, icon transitions, micro-interactions, and performance, with linked recipe files for CSS, Tailwind, and Framer Motion. Use it when building or reviewing UI components, hover states, shadows, or any visual detail where the interface feels subtly off.
• Комплексный аудит интерфейса — better-interface is a Claude Code skill that performs a cross-discipline interface review by routing a screen, flow, feature, or product to six domain skills — better-accessibility, better-layout, better-writing, better-typography, better-colors, and better-ui — then consolidating their findings into a single ranked verdict. The skill owns orchestration only: each domain's rules remain with its owning skill, and better-interface never duplicates or overrides them. Two modes are available — quick (HIGH and MEDIUM issues only, capped at 5 findings along the primary path) and full (up to 15 findings across all states, including empty, loading, error, and narrow-width). Before reviewing, it identifies the framework, styling system, design tokens, and supported viewports so every suggested fix matches the project's own idiom. Use it when a holistic evaluation of a screen or flow is needed rather than a single-domain audit.
• Профессиональная вёрстка и выравнивание UI — better-layout is a Claude Code skill that teaches Claude to structure web UI through grouping, alignment, reading order, and progressive disclosure. It covers spacing between controls, layout margins, adaptive breakpoints, container queries, RTL direction, logical CSS properties, and safe area insets. The skill activates on layout, spacing, alignment, negative space, visual hierarchy, responsive layout, and full-bleed content scenarios. Use it when building or reviewing pages and components for correct spacing and alignment — hit areas and focus behavior are handled by better-accessibility, visual polish by better-ui, and text spacing by better-typography.
• Типографика и оформление текста в UI — better-typography is a Claude Code skill that controls how text renders, wraps, and behaves in the UI — from typeface selection to fine typographic details. It ships seven reference files covering font choosing and pairing, variable fonts and OpenType features, type scale and heading hierarchy, wrapping and punctuation, underlines and selection styling, a CSS cheat sheet with Tailwind equivalents, and a structured review output format with a severity scale. Use it when picking or pairing typefaces, configuring woff2 loading, setting up a type scale, checking heading levels, styling text in components, handling truncation or line clamping, or auditing frontend code for typography issues like broken line-height, excessive letter-spacing, or poor text contrast.
• Психологический аудит текста на резонанс с аудиторией — dbs-resonate is a Claude Code skill that diagnoses content drafts using five communication psychology frameworks to pinpoint exactly why a piece fails to resonate with its audience. It works through a structured workflow: extracting every claim in the draft, identifying the single core mechanism, then evaluating resonance across five theoretical dimensions — Spiral of Silence (Noelle-Neumann, 1974), Uses and Gratifications (Katz, 1959), Framing Theory (Entman, 1993), Two-Step Flow (Lazarsfeld, 1944), and Cognitive Consistency (Festinger, 1957). Each dimension is rated effective, weak, or ineffective with a finding tied directly to specific lines in the draft. The output is an actionable report specifying what to cut, what to compress into supporting detail, what to amplify, and what to keep — plus a one-sentence skeleton of the revised piece. Ideal for video scripts and social posts that feel correct but fail to engage or spread.
• Анализ текста на признаки AI-генерации — dbs-ai-check is a Claude Code skill that scans copy for AI writing fingerprints and produces a numbered detection report with direct quotes from the original text and a severity rating for each signal. It checks 22 characteristics — including suspiciously smooth sentence rhythm, fabricated anecdotes, high-density "not X but Y" reversals, and over-polished emotional arcs — while adjusting thresholds for genre: a closing power line is normal in short-video scripts but a red flag in a long-form essay. By default the skill diagnoses only and does not rewrite; if the user asks to revise, it asks one targeted question per flagged passage to help the author find their own voice rather than swap one AI pattern for another. It is built for copywriters, editors, and content creators who want an honest read on how human their writing actually sounds.
• Доступность интерфейса по стандартам WCAG — better-accessibility is a Claude Code skill that covers accessibility engineering for product interfaces — from focus states and keyboard navigation to ARIA, forms, screen readers, and touch targets following WCAG principles. It organizes guidance into seven areas: focus and keyboard, semantics and ARIA, forms, screen readers, hit areas, motion and zoom, and a structured review output format with a severity scale. Concrete patterns include roving tabindex for composite widgets, inert-based modal focus trapping, aria-invalid with aria-describedby for form errors, prefers-reduced-motion handling, and minimum 24×24 CSS-pixel targets per WCAG 2.5.8 AA. Use it when building or reviewing UI components, modals, menus, or custom widgets, and when a keyboard or screen-reader bug needs to be diagnosed and fixed.
• UX-тексты интерфейса: кнопки, ошибки, подсказки — better-writing is a Claude Code skill that helps write and review any user-facing interface copy: button and link labels, form errors, placeholders, empty states, onboarding flows, settings labels, and notifications. It encodes 11 UX writing principles — one consistent voice with tone that shifts by context, verb-first button labels, error messages that state how to fix the problem adjacent to the failing field, empty states that orient and prompt action, and a unified vocabulary across multi-step flows. Scope is clearly bounded: capitalization via text-transform and truncation belong to the better-typography skill, aria-invalid and live-region markup to better-accessibility, and translated-string spacing to better-layout. Built for developers and designers who write or audit product copy and want interface text that is clear, consistent, and self-explanatory without extra guidance.
• Диагностика контента для соцсетей — dbs-content is a Claude Code skill that diagnoses how to turn a confirmed content topic into high-quality social media content — without writing it for you. It identifies the best format (short video, long video, image-text post, livestream, or article), matches it to the right platform (小红书, 抖音, X, WeChat Official Account), and runs a five-dimension audit: text cleanliness and absence of AI-flavored phrasing, cover and headline effectiveness, expression efficiency, cognitive gap versus competitors, and a concrete AI-assisted workflow recommendation. The session ends with a structured diagnostic report and a single actionable first step. Built for content creators, knowledge bloggers, and founders who monetize through self-publishing and need a clear answer to "how exactly should I execute this topic."
• Генератор заголовков для Xiaohongshu — dbs-xhs-title is a Claude Code skill that selects the best headline formula for Xiaohongshu from a library of 75 validated templates organized across 12 psychological trigger categories, then generates custom titles with a clear explanation of why each formula was chosen. Acting as a formula matcher rather than a freeform title generator, it traces every output to a specific formula number — covering triggers from cognitive conflict and fear of loss to social proof, controversy, and interactive formats. Each recommendation includes the original viral headline from the platform, the formula template, and a one-line rationale. Built-in quality rules cap titles at 20 characters and require at least two of six engagement signals: contrast, numbers, curiosity, conflict, time promise, or result promise. Designed for content marketers, social media managers, and creators who need high-click-rate headlines for the Xiaohongshu platform.
• Ревью качества интерфейса в диффе — interface-review is a Claude Code skill that evaluates interface quality for a specific change — uncommitted work, the current branch, or a pull request — rather than the full codebase. It resolves the change scope, expands affected files to their blast radius of surfaces, reads both sides of every diff hunk, and classifies each finding as `Introduced`, `Regression`, or `Pre-existing`. Removed lines are checked against Removed Signals rules to catch regressions invisible in the post-change state. The final verdict, output format, severity cap, and consolidation rules are handed off to the `better-interface` skill; correctness, tests, security, and performance are explicitly out of scope. Ideal for developers who want automated interface-focused code review that tracks what a change broke, not just what it added.
• Генерация SEO-изображений через Gemini — seo-image-gen is a Claude Code skill that generates production-ready images for SEO use cases through Gemini, powered by the banana Creative Director pipeline via the nanobanana MCP server. It covers eight asset types — OG/social previews (16:9, 1K), blog hero images (16:9, 2K), schema images, social squares, product photography on white backgrounds (4:3, 2K), infographics (2:3, 4K), favicons, and Pinterest pins — each with preset aspect ratios, resolutions, and domain modes applied automatically. Batch generation of multiple variations is supported out of the box. After every successful generation the skill walks through an SEO checklist: keyword-rich alt text, SEO-friendly file naming, WebP conversion targeting under 200 KB, ImageObject schema markup, and Open Graph meta tags. Requires the nanobanana extension installed; built on the open-source Claude Banana project by AgriciDaniel.
• Higgsfield — создание и деплой сайтов и игр — higgsfield-websites is a Claude Code skill that drives the full lifecycle of building, editing, and deploying websites, web apps, and games through the Higgsfield CLI (`higgsfield website …`). It supports three project types selected via `--type` on create: `website` (React 19 + TanStack Start, standalone, no Higgsfield integration or AI generation), `app` (Sign in with Higgsfield, fnf SDK, Quanta UI, D1 auth contract), and `game` (realtime multiplayer browser games built around six pure functions in `logic.js`). Every project runs as a single SSR Cloudflare Worker with D1/R2/KV/DO/Containers and deploys to its own subdomain. Use this skill whenever the task is to launch a landing page, portfolio, SaaS dashboard, AI media generator, or a browser game.
• Генерация SEO-изображений через Gemini — seo-image-gen is a Claude Code skill that generates production-ready images for SEO use cases using Gemini via the nanobanana MCP server. It covers eight asset types: OG/social preview images (16:9, 1K), blog hero images (16:9, 2K), product photography with white studio background (4:3, 2K), infographics (2:3, 4K), schema images, social squares, Pinterest Pins, and favicons — each with pre-configured aspect ratios, resolutions, and Creative Director domain modes from the banana pipeline. After every generation, the skill walks through an SEO checklist covering alt text, keyword-based file naming, WebP conversion via ImageMagick, ImageObject schema markup, and og meta tags. The batch command produces N variations in a single run. Requires the banana extension (nanobanana MCP server) to be installed separately.
• Higgsfield Soul — обучение модели лица — higgsfield-soul-id is a Claude Code skill that trains a personalized Soul identity model on a person's face photos, making it reusable across any number of image and video generations. It accepts 5–20 face photos with varied angles and lighting, and supports two variants: `--soul-2` for image generation and `--soul-cinematic` for cinematic or video work. After training completes, the skill returns a `reference_id` that can be passed to higgsfield-generate using models like `text2image_soul_v2` or `soul_cinema_studio`. A paid Higgsfield Basic+ plan is required. It's the right choice when you need a consistent digital twin or personal avatar across multiple generations — as opposed to one-shot face swaps handled by higgsfield-generate with `--image`.
• SEO-аудит через Google API — seo-google is a Claude Code skill that provides direct access to Google's own SEO APIs: Search Console (Search Analytics, URL Inspection, Sitemaps), PageSpeed Insights v5, CrUX field data with 25-week history, Indexing API v3, and GA4 organic traffic. The skill operates across four credential tiers — from a basic API key (PageSpeed, CrUX, NLP, Knowledge Graph) up to a full setup with a service account and GA4 property ID, which unlocks batch URL inspection, batch indexing of up to 200 URLs, and landing-page reports. Because data comes directly from the Chrome User Experience Report and Search Console rather than a crawler, Core Web Vitals, indexation status, and search positions reflect genuine field measurements. It's built for SEO professionals and developers who need to audit page performance, track CWV trends, submit URLs for indexing, and pull GA4 organic traffic — all from within a single Claude workspace.
• SEO-аудит интернет-магазина и товарных страниц — seo-ecommerce is a Claude Code skill that delivers end-to-end SEO analysis for online stores, covering product page optimization, schema validation, competitor pricing intelligence, and keyword gap discovery between organic search and Google Shopping. It operates in two modes: standalone for on-page audits and schema checks without any external API, and enhanced via the DataForSEO Merchant API for live Google Shopping and Amazon data. Four commands are available: `/seo ecommerce <url>` for a full product page audit, `/seo ecommerce products <keyword>` for Shopping competitor analysis, `/seo ecommerce gaps <domain>` to surface keyword gaps, and `/seo ecommerce schema <url>` for schema markup validation. Every paid Merchant API call is gated by a mandatory cost check that requires explicit user approval before proceeding. Built for e-commerce SEO specialists and store owners who need actionable data across both search results and marketplace listings.
• Higgsfield — брендовые фотосессии продуктов — higgsfield-product-photoshoot is a Claude Code skill that generates brand-quality product images using the Higgsfield CLI with backend prompt enhancement and final rendering via gpt_image_2. It supports 10 modes: product_shot, lifestyle_scene, closeup_product_with_person, moodboard_pin, hero_banner, social_carousel, ad_creative_pack, virtual_model_tryout, conceptual_product, and restyle — covering studio catalog shots, in-context lifestyle scenes, Pinterest pins, wide-format hero banners, coordinated ad packs for Meta and TikTok, virtual try-on, and surreal CGI visuals. Before submitting, the skill asks up to four short questions about count, style, and intended platform; the backend assembles the final prompt automatically. Designed for e-commerce product photography, paid social creatives, landing pages, and brand campaigns.
• Higgsfield AI — генерация медиаконтента — higgsfield-generate is a Claude Code skill that submits image, video, 3D asset, and audio generation jobs to Higgsfield AI via the `higgsfield` CLI. Default models include GPT Image 2 for high-fidelity images, graphic design, and on-image text; Seedance 2.0 for cinematic video up to 4K (4–15 s, including image-to-video); Nano Banana 2/Lite/Pro for character and stylized work; Marketing Studio for ads, UGC, unboxing, and presenter videos; and Seed Audio 1.0 for text-to-audio, sound effects, and music. The skill also covers video editing and remixing, 3D/GLB model creation, workflow execution, and Virality Predictor scoring for finished videos. It is the go-to choice for developers and marketers who need to automate media production from a text prompt or reference image all the way to a polished video or ad.

FAQ

How do I make Claude write in my brand voice?

Voice is set by examples and prohibitions, not adjectives. A skill describes the structure, three to five sample paragraphs that worked, and an explicit list of what must not appear. That holds up far better than "write in a friendly but professional tone", which gets reinterpreted on every run.

How is a content skill different from a saved prompt?

A prompt is a one-off you paste and adjust each time. A skill sits in the project and loads itself when the task arises, so its rules reach every piece of copy in the project — not only the ones where somebody remembered the right prompt.

All Claude Code skills